CRISIS AND NETWORKS CONGESTION PREVENTION:
THE GAME OF DETERRENCE APPROACH
ORT France
Free traffic flow plays an important role in normal life of developed societies. It may become a core issue in periods of crisis. Such would be the case for instance for civil crises generated by earthquakes, floods, tsunamis, or heavy fires (like the 2009 Australian bushfires). But it will be also the case for crises stemming from military or terrorist issues. In both types of crises, a variety of networks may be affected like transport (road, railways, air, boats), communication, water, electricity etc., with possibly dramatic consequences. Thus congestion in road transport due to destruction of infrastructures may prevent teams to rescue the victims in required time, or to send reinforcements to back up troops facing an enemy offensive. Therefore availability of congestion prevention systems enables governments, local authorities but also private entities to dramatically increase crisis management efficiency.
In that respect the present paper proposes a general method to prevent congestion in point to point networks, and support a serious game enabling to develop skills and competencies in networks infrastructures and traffic management. Beyond theoretical results, the proof of concept is given on the case of road traffic, for whicha number of simulation tools have been developed like PARAMICS, CORFLO, CORSIM, or CONTRAM [see in particular 6,7], many based on queuing theorywith problems of computational performance, model accuracy or integrating management and information due to the network size.
The basic assumption is that congestion prevention requires sensors and traffic regulation components which can be of different kinds, like military or police personnel or traffic lights. In the following we shall consider one kind, i.e. traffic lights, and distinguish between four levels of analysis corresponding to a relation of inclusion between the four associated sets:
- traffic light
- crossroad
- regional traffic network
- global traffic network.
- reception: receive vehicles from adjacent nodes or from outside the network
- emission: send vehicles to adjacent nodes or outside the network.
Thus a crossroad may include several traffic lights while a regional traffic network will
include several crossroads, and likewise a global traffic network will include several
regional traffic networks.
The proposed approach combines general principles used for preventing congestion in point to point telecommunication networks with Games of Deterrence. These games originally developed for strategic analysis, have proven to have applications in a variety of fields like business process reengineering [1], traffic control in telecommunication networks [2], inference schemes and reliability [9], or multi-criteria decision making [11].
The starting point is given by two common sense principles:
– on the one hand given that congestion results from a level of the incoming traffic that cannot be sustained by the network in its present state of operations, incoming traffic should be stopped before the network is overwhelmed
– on the other hand, traffic can be modelled through a network, each node of which has two possible functions:
Each node can be controlled by a set of traffic lights depending on the complexity of traffic rules concerning that node. These traffic lights will be used to implement the solutions of the
traffic control system as given by the solutions of the associated game of deterrence. More precisely, in a first part, after defining congestion, we shall analyze the role of traffic lights and crossroads as instruments for traffic control.
Then starting from elementary examples of traffic control issues dealing with motorway access ramp, or narrowing of a road, we shall gradually increase the complexity of the networks being considered, through enabling turns and increasing the number of 32 crossroads. A matrix representation of each network state will then be proposed,
A second part will recall the main concepts and properties of matrix games of deterrence.
In a third part we shall show how road traffic networks including crossroads equipped with traffic lights can be modelled as finite state machines, which states can in turn be represented by matrix games of deterrence.
A fourth part will present an algorithm based on games of deterrence solutions, and that will prevent congestion within the network.
A fifth part will depict simulations that have been performed in order to bring the experimental proof of the algorithm validity, and hence the proof of concept.
A sixth and last part will address the issue of how to make sure that congestion prevention within the network doesn’t translate into generating congestion outside the network. To that end, the approach will consider a global network that will be broken down into regional ones.